AArch64 error 200108231 at higher optimization levels
Original Reporter info from Mantis: crlab @neurolabusc1
-
Reporter name: Chris Rorden
Original Reporter info from Mantis: crlab @neurolabusc1
- Reporter name: Chris Rorden
Description:
Lazarus now builds for AArch64, however, setting the Build Options for "Optimized IDE" invokes the optimized -O3 option. This generates the an error 200108231. Presumably, this is related to issue 0034786 reported for the Raspberry Pi (also an ARM architecture). So maybe solutions for that issue can be applied here. The unit that fails to compile is grapmhath.pp that does use extended floating point a lot, and as an aside changing this to the double format that is natively supported by modern CPUs seems a nice idea. However, changing "extended" to "double" does not resolve the error.
I have included a simple freepascal command line project that replicates the error without having to include Lazarus...
Steps to reproduce:
fpc -O3 ax.pas
...
graphmath.pp(518,7) Fatal: Internal error 200108231
Mantis conversion info:
- Mantis ID: 37397
- OS: Darwin
- OS Build: 10.11.6
- Build: 45816
- Platform: MacBook 2012 Retina 13"
- Version: 3.3.1
- Fixed in version: 3.3.1
- Fixed in revision: 45823 (#dd6e6eb2)